网络GIS中最佳负载均衡的分布式缓存副本策略

云环境下的网络地理信息服务具有分布性和异构性,空间数据(瓦片)的访问请求具有高度聚集性和不均匀性。以最小化负载不均衡度为目标,提出了一种应用于异构的、分布式高速缓存集群系统的多副本策略。该策略针对瓦片访问请求存在不均衡性,最小化热点访问数据的通信权重值,最大化地利用分布式集群缓存能力生成副本;针对异构集群环境下服务器处理能力的不均衡性,根据服务器性能和瓦片副本的通信权重值,匹配各个服务器的缓存能力部署副本。实验证明,该策略避免服务器拥塞的同时,能充分利用有限的分布式集群缓存能力,实现较好的负载均衡和较高的资源利用率,并能获得良好的缓存命中和请求响应性能。

A Replication Strategy Based on Optimal Load Balancing for a Heterogeneous Distributed Caching System in Networked GISs

Networked geospatial information services in cloud-based environments are distributed and heterogeneous; accesses to geospatial data (tiles) are uneven and has the feature of a high degree of aggregation. Aiming to minimize the degree of load imbalance, this paper proposes a replication strategy for a heterogeneous distributed high-speed cluster-based caching system. First, taking into account the unbalanced accesses to tiles, it minimizes the weighted communication values of hotspot tiles, and generates the maximum number of replicas based on the total cache capability of the distributed cluster-based caching system. Then, since each server has a different processing capacity in the heterogeneous system, the strategy places the replicas based on the service performance of each caching server and the weighted communication value of each replica, thus matching the cache capacity of each server. Experimental results reveal that the proposed strategy can avoid server congestion while fully utilizing limited cache capacity to achieve a better load balancing and a high resource utilization, delivering good response performance and a high cache hit rate.